stable oncogenic silencing in vivo by programmable and ......stable oncogenic silencing in vivo by...

OPEN

ORIGINAL ARTICLE

Stable oncogenic silencing in vivo by programmable andtargeted de novo DNA methylation in breast cancerS Stolzenburg1,2, AS Beltran2, T Swift-Scanlan3, AG Rivenbark4, R Rashwan1 and P Blancafort1,2

With the recent comprehensive mapping of cancer genomes, there is now a need for functional approaches to edit the aberrantepigenetic state of key cancer drivers to reprogram the epi-pathology of the disease. In this study we utilized a programmableDNA-binding methyltransferase to induce targeted incorporation of DNA methylation (DNAme) in the SOX2 oncogene in breastcancer through a six zinc finger (ZF) protein linked to DNA methyltransferase 3A (ZF-DNMT3A). We demonstrated long-lastingoncogenic repression, which was maintained even after suppression of ZF-DNMT3A expression in tumor cells. The de novo DNAmewas faithfully propagated and maintained through cell generations even after the suppression of the expression of the chimericmethyltransferase in the tumor cells. Xenograft studies in NUDE mice demonstrated stable SOX2 repression and long-term breasttumor growth inhibition, which lasted for 4100 days post implantation of the tumor cells in mice. This was accompanied with afaithful maintenance of DNAme in the breast cancer implants. In contrast, downregulation of SOX2 by ZF domains engineered withthe Krueppel-associated box repressor domain resulted in a transient and reversible suppression of oncogenic gene expression. Ourresults indicated that targeted de novo DNAme of the SOX2 oncogenic promoter was sufficient to induce long-lasting epigeneticsilencing, which was not only maintained during cell division but also significantly delayed the tumorigenic phenotype of cancercells in vivo, even in the absence of treatment. Here, we outline a genome-based targeting approach to long-lasting tumor growthinhibition with potential applicability to many other oncogenic drivers that are currently refractory to drug design.

Oncogene (2015) 34, 5427–5435; doi:10.1038/onc.2014.470; published online 16 February 2015

INTRODUCTIONCytosine methylation in mammalian DNA is regarded as a keyepigenetic modification controlling essential processes such asimprinting, silencing of retrotransposons and cell differentiation.1

In addition to DNA methylation (DNAme), repressive histone post-translational modifications reinforce gene inactivation, resulting inthe formation of inactive chromatin, namely heterochromatin.2

Importantly, the epigenetic information must be stably trans-mitted during mitosis for the proper maintenance of cellularidentity, a process referred as epigenetic memory.3 In pathologicalstates such as cancer, the epigenetic landscape of normal cellsbecomes disrupted. Aberrant incorporation or removal of DNAmein cancer cells leads to genome-wide alteration of geneexpression, including inactivation of tumor suppressor genesand reactivation of oncogenes.4 Notably, changes in DNAmepatterns are characteristic hallmarks of the distinct intrinsicsubtypes of breast cancer. Poorly differentiated, highly prolifera-tive basal-like breast cancers associated with stem/progenitor cell-like features are significantly hypomethylated relative to the otherbreast cancer subtypes. Importantly, these tumors are driven byaberrant activation of multiple developmental transcriptionfactors (TFs), which fuel the tumor with sustained proliferation,drug resistance and metastatic capacity.5–7

The High Mobility Group oncogenic TF SOX2 is normallyexpressed in embryonic stem cells and neural progenitor cells,

where it maintains self-renewal.8,9 DNAme in the SOX2 promoterand enhancer regions functions as an epigenetic switch, whichforces cells to activate multiple differentiation pathways.10 SOX2 istherefore not expressed in most normal adult tissues.10–12

Moreover, aberrant reactivation of SOX2 has been detected in~ 43% of basal-like breast cancers and in several other malig-nancies, including glioblastoma, lung, skin, prostate and ovariancarcinomas.13 SOX2 overexpression in tumor specimens has beenassociated with both promoter hypomethylation relative toadjacent normal tissue and copy number amplifications.14,15 Theoverexpression of SOX2 in breast cancer has been shown todirectly activate CYCLIN D1, resulting in an increased mitotic indexand proliferation.16,17 The downregulation of SOX2 by RNAinterference decreased the tumorigenic phenotype in the lung,breast and ovarian cancers.13,16,18 However, a major limitation ofsmall interferin RNA and small hairpin RNA approaches in cancertherapy has been the short half-life of small RNA or themethylation of the virally encoded small hairpin RNA promoter,which results in transient effects in vivo.In light of the essential role of DNAme in the regulation of SOX2

expression we hypothesized that targeted de novo methylation inthe SOX2 promoter would result in an epigenetic 'off' switch,forcing cancer cells to undertake differentiation programs.Furthermore, because DNAme is read and written by endogenousproteins and faithfully transmitted during cell division, we

1Cancer Epigenetics Group, The Harry Perkins Institute of Medical Research, Western Australia & School of Anatomy, Physiology and Human Biology, M309, The University ofWestern Australia, Nedlands, Western Australia, Australia; 2Department of Pharmacology, School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA;3Lineberger Comprehensive Cancer Center/School of Nursing, University of North Carolina, Chapel Hill, NC, USA and 4Department of Pathology and Laboratory Medicine,University of North Carolina School of Medicine, Chapel Hill, NC, USA. Correspondence: Professor P Blancafort, Cancer Epigenetics Group, The Harry Perkins Institute of MedicalResearch, Western Australia & School of Anatomy, Physiology and Human Biology, M309, The University of Western Australia, 6 Verdun Street Nedlands, Nedlands, WesternAustralia 6009, Australia.E-mail: [email protected] 14 July 2014; revised 15 November 2014; accepted 9 December 2014; published online 16 February 2015

Oncogene (2015) 34, 5427–5435© 2015 Macmillan Publishers Limited All rights reserved 0950-9232/15

www.nature.com/onc

http://dx.doi.org/10.1038/onc.2014.470

mailto:[email protected]

http://www.nature.com/onc

reasoned that artificial incorporation of de novo DNAme in theSOX2 promoter would confer stable oncogenic silencing, resultingin a sustained blockade of cell growth, faithfully propagated insuccessive cell generations. The forced epigenetic reprogrammingof an oncogene toward a 'normal-like' state, which will be stablytransmitted through mitosis, remains a hitherto unexploredparadigm in cancer research; the resulting therapeutic effectwould provide durability and sustained anticancer response(memory).Herein we took advantage of previously characterized artificial

TFs targeting the SOX2 promoter made of six zinc finger (6ZF)domains, recognizing 18 base pair (bp) sites in the corepromoter.17 These 6ZFs were linked to the catalytic domain ofDNA methyltransferase 3A (DNMT3A), an enzyme that catalyzes denovo DNAme, to correct the aberrant methylation state of SOX2 incancer cells. We show that targeted DNAme is sufficient to initiate,and even reinforce, potent and mitotically heritable silencing ofSOX2 expression. Our results indicate that the engineering ofartificial DNA-binding domains (DBDs) linked to DNMT3A can beutilized to promote long-lasting SOX2 silencing and represents apromising therapeutic approach to minimize tumor relapse. Ourapproach could be used to promote customizable heterochroma-tization and epigenetic silencing of elusive cancer drivers mappedby recent genomic projects, such as TFs and small GTPAses (RAS),for which no drug is currently available.

RESULTSZF598-DNMT3A initiates SOX2 downregulation accompanied bystable inhibition of cancer cell growthTo induce targeted DNAme in the SOX2 locus we took advantageof our previously characterized 6ZF domains designed torecognize 18 bps sequences in the SOX2 proximal promoter(Figure 1a).17 Subsequently, we engineered the catalytic domainof the human DNMT3A to the C terminus of the 6ZF domainsand in frame with a 13 amino-acid flexible linker, to producethe ZF598-DNMT3A (Figure 1a) and the ZF552-DNMT3A(Supplementary Figure S1) constructs.19 The same 6ZF arrayslinked to the repressor super Krueppel-associated box domain(SKD, ZF598-SKD) were used as a control, as this domain promotespotent KAP1-dependent repression when linked at the N terminusof the 6ZFs, without induction of DNAme.17,19

We chose the MCF7 breast cancer cells as a model cell line tostudy the temporal dynamics of the incorporation of DNAme, asthese cells express high levels of SOX2, and the promoter does notcontain methylated CpG dinucleotides (Figure 2a). The 6ZFconstructs were delivered into MCF7 cells using inducible retro-viral vectors17 and stable clones were isolated in which thetemporal expression of the ZF598-DNMT3A and ZF598-SKDfusions was controlled by doxycycline (Dox) treatment.Upon induction of the ZF598-DNMT3A and ZF598-SKD expres-

sion with Dox (+Dox), SOX2 mRNA levels decreased by 90% and73%, respectively, as compared with control (empty vector)transduced cells (Figure 1b). When Dox was removed for up to8 days (~10 cell generations) from the culture media (R, Figures 1band c) cells transduced with ZF598-SKD restored SOX2 expressionto levels that were similar to those of uninduced cells, whereascells expressing the ZF598-DNMT3A showed a persistent decreaseof SOX2 expression (85% relative to uninduced cells) upon Doxremoval (Figure 1b). These findings suggested that the ZF598-DNMT3A construct conferred transcriptional memory in thetargeted SOX2 locus. The SOX2 transcriptional silencing inducedby ZF598-DNMT3A was accompanied by a significant decrease inSOX2 protein expression. This suppression was sustained in Dox-removal conditions in the absence of expression of the ZF598-DNMT3A, and even resulted in increased levels of SOX2 silencingrelative to +Dox cells as demonstrated by western blot (Figure 1c).

We next investigated whether the ZF598-DNMT3A fusion wasable to maintain suppression of tumor cell growth in Dox-removalconditions. We monitored changes in cell viability over timeassociated with induction and subsequent removal of the ZFprotein expression (Figure 1d). MCF7 cells stably transduced witheither empty vector control, ZF598-SKD or ZF598-DNMT3A, weretreated for 72 h with Dox to induce the expression of the ZFproteins. Next, Dox was removed from the culture media(Figure 1d, red arrow) and cell viability was monitored for thenext 144 h. Cells expressing empty vector or ZF598-SKD hadsignificantly higher proliferation rates upon Dox removal ascompared with ZF598-DNMT3A transduced cells, which retaineda more robust inhibition of cell proliferation after Dox removal.These results demonstrate the unique capacity of the 6ZF-DNMT3A fusions to establish an oncogenic silencing state at boththe mRNA and protein level that was stably maintained throughcell generations.

ZF598-DNMT3A silences SOX2 by initiation and faithfulpropagation of DNAmeWe subsequently investigated whether ZF598-DNMT3A was ableto catalyze targeted de novo DNAme in the SOX2 promoter. MCF7cells stably transduced with ZF598-DNMT3A were treated withDox to induce the expression of the ZF598-DNMT3A. Next, cellswere either harvested at 72 h or removed from Dox and passagedfor 8 additional days in Dox-free conditions. These Dox-removalconditions completely suppressed the expression of the ZFprotein, as determined by western blotting of nuclear extractsusing an anti-HA antibody (Figure 1c).Conventional sodium bisulfite sequencing was first conducted

to quantify the incorporation of DNAme in the amplicon contain-ing the ZF598-DNMT3A binding site (Amplicon II, Figure 1a). Dox-induced expression of ZF598-DNMT3A resulted in an increase ofDNAme of up to 90% at specific CpG dinucleotides (Figure 2a). Theincorporation of DNAme was also validated by MassARRAYanalysis (Figures 2b and d). Control cells or cells expressing thecatalytic mutant ZF598-DNMT3A-E74A revealed no increase ofCpG methylation, demonstrating that the induction of DNAmerequired the catalytic activity linked to the ZF598. Similar resultswere obtained utilizing a second ZF protein (ZF552-DNMT3A)targeting an 18 bps region 552 bps upstream of the translationstart site (Supplementary Figure S1), which demonstratesthat the approach is generalizable for other DBDs targetingSOX2. Furthermore, ectopic overexpression of the catalytic domainof DNMT3A (untargeted DNMT3A) lacking the SOX2-specific DBDsresulted in an unspecific background incorporation of DNAme(Supplementary Figure S1). The untargeted construct was notcapable of downregulating SOX2 messenger RNA levels or toinduce proliferative arrest. Furthermore, when the DNMT3A waslinked to 6ZF DBDs specific for the MASPIN tumor suppressor genecontext, instead of an oncogenic context, targeted methylationwas observed in the MASPIN promoter accompanied by anincrease in cell proliferation.19 Reciprocally, no changes in DNAmewere observed in the MASPIN promoter context with the SOX2-specific DBDs linked to DNMT3A (Supplementary Figure S2). Theseresults indicated that the approach required two functionalcomponents, the targeting DBD and a catalytically active DNMT3Adomain.Importantly, the de novo pattern of CpG methylation initiated

by ZF598-DNMT3A in the SOX2 promoter was stably and faithfullymaintained over several cell generations in vitro, even 8 days afterDox removal (Figure 2b) when the expression of the targetedmethyltransferase was not longer detectable by immunoblotting.Furthermore, both the pattern and intensity of the inducedmethylation (40–90% depending on the specific CpG nucleotide)were maintained over the time course of the experiment.

Programmable oncogenic silencing in breast cancerS Stolzenburg et al

5428

Oncogene (2015) 5427 – 5435 © 2015 Macmillan Publishers Limited

To evaluate the spreading of DNAme in the SOX2 locus after theinitial induction, we analyzed two additional amplicons (ampliconI and III) flanking the 6ZF-binding site, encompassing ~ 1 kbps up-and downstream of the translation start site (Figures 2c and d).Sequencing analysis of amplicon I (−1069 to − 623 bps upstreamof the translation start site) demonstrated an increase of DNAmeupon ZF598-DNMT3A expression as compared with the MCF7empty vector transduced cell line or ZF598-DNMT3A-E74A-transduced cells. The frequencies of DNAme in amplicon I weresignificantly higher than those immediately adjacent to the ZF-binding site, suggesting that DNAme was possibly spread andeven reinforced in the flanking regions. Most importantly, afterdiscontinuation of the ZF598-DNMT3A expression (Dox removal),the de novo methylation additionally increased up to 97% atspecific CpG sites in Amplicon I (Figure 1c).As last, the MassARRAY analysis of DNAme in amplicon III (+695

to +1055 bps downstream of the translation start site, Figure 2d)revealed no methylation in MCF7 control and ZF598-DNMT3A cellsin the absence of Dox. Upon ZF598-DNMT3A expression (+Dox)CpG methylation increased up to 80%, and no DNAme wasinduced in cells expressing the ZF598-DNMT3A-E74A mutant. Ourtime-course analysis suggests that the de novo DNAme induced bythe artificial methyltransferase results in both, a phase of inductionof gene silencing upon Dox induction, and a phase ofmaintenance and reinforcement of silencing (Dox removal), which

could be associated with propagation or spreading of DNAmefurther away from the 6ZF site during DNA replication. Theseresults further support the increased downregulation of SOX2expression after Dox removal previously observed by quantitativereverse transcription polymerase chain reaction and western blotanalysis and outline the importance of CpG dinucleotides flankingthe core promoter for the regulation of SOX2 expression.

ZF598-DNMT3A expression confers anticancer memory andreduces tumor growth in a breast cancer xenograft in NUDE miceTo analyze whether the ZF598-DNMT3A construct inducedphenotypic memory in vivo we took advantage of our inducibleMCF7 cell lines stably transduced with either the catalyticallyactive ZF598-DNMT3A or the empty vector control. A total of2 × 106 MCF7 cells transduced with either ZF598-DNMT3A orcontrol were implanted into the flank of NUDE mice and allowedto grow for 22 days before switching the animals to a Dox-containing diet (Figure 3a). Within each group N= 5 animals weremaintained in a Dox-free diet. At day 19 post induction, half of theZF598-DNMT3A injected animals from the +Dox group wereremoved from the Dox diet (N= 10), to withdraw the expression ofthe ZF methyltransferase (R). In contrast with ZF598-DNMT3A,catalytic dead ZF598-DNMT3A-E74A mutant cells failed to reduce

empt

y vec

tor

empt

y vec

tor

ZF598-

SKD

ZF598-

SKD

ZF598-

SKD

ZF598-

DNMT3A

ZF598-

DNMT3A

ZF598-

DNMT3A

0.0

0.5

1.0

1.5

DOX - + - + R - + R

rel.S

OX

2 m

RN

Aex

pre

ssio

n

******

0 48 960

5

10

15 empty vector

ZF598-DNMT3A

ZF598-SKD

time (hrs)

Fo

ld in

crea

se in

AT

P r

elea

se r

el. t

o t

=0

5’3’

3’5’

-654 -279

-1069 -623 +695 +1055ATG

GCCCCCTCCTCCCCCGGC

III

III

ZF

linker

DNMT3A

144

empty ve

ctor

empty ve

ctor

ZF598-SKD

ZF598-SKD

ZF598-DNMT3A

ZF598-DNMT3A

ZF598-DNMT3A

ZF598-SKD

DOX - + - + R - + R

SOX2

HA-tag

HA-tag

H3

37kDA

42kDA

55kDA

17kDA

Figure 1. Stable downregulation of the SOX2 expression by ZF598-DNMT3A. (a) Schematic illustration of the SOX2 promoter indicating thedomain structure of the ZF598-DNMT3A construct, its binding site and the three amplicons analyzed by sodium bisulfite sequencing orMassARRAYs (amplicon I, blue; II, purple and III, green). (b) Quantification of SOX2 mRNA expression by qRT-PCR in MCF7 cells. Cells werestably transduced with empty vector, ZF598-SKD and ZF598-DNMT3A. The expression of the ZF fusion was controlled by addition orremoval of doxycycline (Dox); R=Dox removal. Cells were induced with Dox every 48 h and either harvested at 72 h after first induction(+Dox) or removed from Dox and subcultured for 8 days and processed for western blot. Error bars represent standard deviation (s.d.)(***Po 0.001, **Po0.01, *Po0.05). (c) Detection of SOX2 by western blot. The C-terminal Hemagglutinin (HA) tag was used forimmunodetection of the ZF proteins. An anti-histone H3 antibody was used as loading control. (d) Cell viability of MCF7 cells assessed by CellTiterGlo assays. Empty vector, ZF598-SKD and ZF598-DNMT3A-transduced cells were induced with Dox for 48 h and removed from Dox after72 h (red arrow). P-values between empty vector and ZF598-DNMT3A and between ZF598-SKD and ZF598-DNMT3A were Po0.001 andPo0.05 respectively, at 144 h.


5429

© 2015 Macmillan Publishers Limited Oncogene (2015) 5427 – 5435

+1055

ZF598-DNMT3A

ZF598-DNMT3A E74A

MCF7

-654 -279

Amplicon II

Amplicon II

Amplicon I

no Dox

Binding site

+Dox

no Dox

+Dox

no Dox

+Dox

-279-654

(R)

(R)

-1069 -623

+695 Amplicon III

(R)

Figure 2. Expression of the ZF598-DNMT3A induces targeted DNA methylation in the SOX2 promoter. (a) Sodium bisulfite-sequencing analysisof DNA derived from MCF7 cells stably transduced with empty vector, ZF598-DNMT3A and ZF598-DNMT3A-E74A mutant and induced withDox. The analyzed amplicon II expands from − 654 to − 279 base pairs (bps) upstream the translation start site and includes the 6ZF-bindingsite (position − 598 relative to the translation start site). Gray circles indicate not analyzed methylation values owing to CpGs with high- or low-mass Dalton peaks falling outside the conservative window of reliable detection for the EpiTYPER software, colored circles indicate variouslymethylated CpGs. (b) MassARRAY analysis of the same amplicon as in (a). Circles indicate the CpG dinucleotides in the amplicon (Color code:yellow=unmethylated CpG to blue= 100% methylated CpG). The percentage of meCpG was determined in MCF7 control-transduced cellsand in MCF7 cells stably transduced with ZF598-DNMT3A or ZF-598-DNMT3A-E74A in absence of Dox (no Dox), presence of Dox (+Dox) orafter Dox removal (R, eight days). (c) MassARRAY analysis of amplicon I (−1069 to − 623 bps upstream of the translation start site) in Dox-freeconditions (no Dox) and after Dox induction (+Dox) Dox removal (R). (d) MassARRAY analysis of amplicon III (+695 to +1055 bps) downstreamof the translation start site, same conditions as above.


5430


MCF7 cell viability in vitro and thereby were not further injected inmice (Supplementary Figure S3).Over the period of 43 days, a significant inhibition (P= 0.0001)

of tumor growth was detected in animals injected with ZF598-DNMT3A receiving a Dox-containing diet (Figure 3b, right panel),which was superior to the inhibition we have reported forKrueppel-associated box-containing repressors.17 Furthermore,this reduction of tumor growth was maintained up to 72 daysafter Dox induction (P= 0.001). Importantly, two animals inducedwith the ZF598-DNMT3A construct had completely regressedtumor burden and could not be detected by caliper measure-ments. As expected, no significant change in tumor growth wasdetected in animals injected with empty vector (Figures 3b and c,left panel).To study the stability of the therapeutic effect we removed the

Dox from the diet of half (N= 10) of the ZF598-DNMT3A animals

(Dox-removal group), while maintaining the other half under +Doxconditions. These groups allowed us to investigate whether thetumor growth inhibition mediated by SOX2 methylation wasstably transmitted even after removal of ZF598-DNMT3A expres-sion. Tumor volumes in the Dox-free, Dox-induced and Dox-removal groups were monitored for 43 days post induction, equalto 24 days post removal of Dox for the Dox-removal group(Figure 3c, right panel). The tumor volumes of ZF598-DNMT3A+Dox animals demonstrated a significant inhibition relative toZF598-DNMT3A −Dox animals (P= 0.0001). Furthermore, miceremoved from a Dox diet maintained a significant reduction oftumor burden (P= 0.004) relative to uninduced animals (Figure 3c,right panel). It should be noted that the tumor sizes of miceremoved from Dox slowly increased over time when comparedwith animals continuously placed under +Dox conditions. How-ever, this effect could be due to the positive selection of cells

724329190-22 64

Implantation

Dox-induction

Dox-removal (R)

sampling+/-Dox (29 days)R (10 days)

sampling+Dox (43 days)R (24 days)



days100

0 7 19 29

Tum

or

volu

me

mm

3

0

50

100

150

Empty vector +DoxEmpty vector -Dox

Days post inductionTu

mo

r vo

lum

e m

m3

0

50

100

150

Days post induction

0 7 19 29 43 64 72

ZF598-DNMT3A -DoxZF598-DNMT3A +DoxZF598-DNMT3A Dox-removal

Dox-removal

Tum

or

volu

me

mm

3

Day 29 post induction

0

50

100

150

emptyvecto

r

emptyvecto

r

ZF598-DNMT3A

ZF598-DNMT3A

Dox - + - +

p=0.0001

Tum

or

volu

me

mm

3

Day 43 post induction

0

50

100

150

ZF598-DNMT3A

ZF598-DNMT3A

ZF598-DNMT3A

Dox - + R

p=0.0001 p=0.004

p=0.0029

Figure 3. The de novo DNA methylation and oncogenic silencing induced by ZF598-DNMT3A are maintained long term in a xenograft modelof breast cancer. (a) Timeline of the subcutaneous tumor injections of MCF7 cells in NUDE mice. (b) Time course plot monitoring tumorvolumes of empty vector control and ZF598-DNMT3A animals induced (+Dox) and uninduced (−Dox). Left panel: tumor growth of emptyvector control animals. Right panel: tumor volumes of animals implanted with ZF598-DNMT3A. (c) Left panel: tumor volumes of ZF598-DNMT3A and control animals at day 29 post induction. Right panel: tumor volumes of ZF598-DNMT3A implanted animals at day 43, when the−Dox control tumors were collected. P-values between groups are indicated.


5431


carrying low levels of methylation in the xenografts or byfunctional compensation by another oncogenic driver.To verify that the reduction of tumor growth in the ZF598-

DNMT3A +Dox animals was associated with the incorporation ofDNAme, MassARRAY analysis of tumor DNA was performed atdifferent time points post induction (Figure 4). DNA samplesextracted from empty vector +Dox tumors at 29 days post inductionwere used as control. Upon induction of ZF598-DNMT3A expressionan increase of DNAme was detected at all time points of samplingrelative to control. The comparison of meCpG frequencies in theSOX2 amplicon I at day 43 post induction indicated that the majorityof CpG dinucleotides were significantly more methylated in theZF598-DNMT3A +Dox and Dox-removal tumors over the ZF598-DNMT3A −Dox tumors sampled in the experiment (SupplementaryFigure S4 and Supplementary Table S1). Importantly, thismethylation was maintained in vivo for more than 50 days afterDox removal in all samples analyzed. These data clearly demon-strated that the targeted DNAme was associated with tumor growthinhibition and with a significant decrease in the rate of tumorrelapse when the treatment was discontinued.

Breast tumor growth inhibition was maintained upon clearance ofZF598-DNMT3A expressionTo examine changes in the tumor morphology, Hematoxylin andEosin staining was performed on sections of empty vector controland ZF598-DNMT3A tumors. The control tumors were collected29 days post induction, the ZF598-DNMT3A tumors werecontinuously induced with Dox for 72 days and for the ZF598-DNMT3A Dox-removal tumors, Dox was withdrawn for 53 daysafter an initial induction of 19 days. Histological analysis of emptyvector +Dox tumor sections revealed a high density of closelypacked tumor cells at day 29 post induction (Figure 5a, left panel).In contrast, the ZF598-DNMT3A +Dox tumors exhibited a verystriking change in the tumor architecture, with a more organizedstructure consisting of islands of tumor cells and an increase ofintervening stroma, a phenotype that was maintained afterremoval of ZF598-DNMT3A expression (Figure 5a). In the analyzedtumors, this change in the tumor phenotype was accompaniedwith the significant downregulation of some mesenchymalmarkers, such as SOX2, TWIST1 and Vimentin and a significantupregulation of some epithelial junction proteins such as Claudin 4as assessed by quantitative reverse transcription polymerase chainreaction (Supplementary Figure S5).

Immunofluorescence analysis of tumor sections revealednuclear expression of the ZF598-DNMT3A protein in the +Doxgroup, but no observable signal in ZF598-DNMT3A −Dox animals(Figure 5b). This induction of methyltransferase expressioncorrelated with a significant decrease in SOX2 expression in thetumors that received Dox, which was not observed in control(empty vector ±Dox) and ZF598-DNMT3A −Dox tumors. Afterremoval of Dox, the ZF598-DNMT3A expression was not longerdetected in the tumor sections. Importantly, a decrease of SOX2expression was stably maintained in vivo after Dox removalrelative to control or −Dox tumors. In addition, the down-regulation of SOX2 expression correlated with a decrease in tumorcell proliferation, as indicated by the Ki-67 marker, which wasmaintained downregulated relative to uninduced cells even afterremoval of ZF598-DNMT3A expression for 10 days (Figure 5b).These results suggest that the de novo methylation patterns wererobustly maintained during somatic cell division in vivo resultingin a remarkable phenotypic change of the tumor cells, suggestingloss of some mesenchymal features and gain of some epithelial-like features.

DISCUSSIONPromoter CpG methylation has an important role in controllinggene transcription and therefore contributes to the regulation ofmany biological processes. In cancer, aberrant DNAme isassociated with initiation and progression of malignant disease.Like TFs, many oncogenic drivers in cancer are of undruggablenature, such as the small GTPases KRAS and HRAS. Therefore, theability to program a heritable targeted silencing state in suchmajor oncogenic drivers would be a high impact accomplishmentwith far reaching clinical implications for cancer treatment.Targeted DNAme by designer ZFs linked to the catalytic domain

of DNA methyltransferases has been recently demonstrated byour group and others.19–25 We have previously shown thatengineered 6ZFs binding the MASPIN tumor suppressor promoterlinked to DNMT3A induced targeted DNAme in the promoter, andthat the resulting phenotypic effect of these proteins was theactivation of cancer cell growth.19 Here, we demonstrated thatDNAme targeted to the SOX2 oncogenic promoter via 6ZFdomains linked to DNMT3A was associated with a significantinhibition of tumor growth in a xenograft mouse model of breastcancer. We took advantage of two different 6ZF proteinsengineered to bind SOX2 promoter (ZF598 and ZF552)26 and

Empty vector

ZF598-DNMT3A

Dox removal (a)

ZF598-DNMT3A

Dox removal (b)

ZF598-DNMT3A

Dox removal (c)

ZF598-DNMT3A

Dox removal (d)

Dox removal (e)

Day 29post-induction





Figure 4. MassARRAY analysis of tumor DNA shows targeted induction and maintenance of DNA methylation. DNA from ZF598-DNMT3A+Dox and Dox-removal tumors was extracted at the indicated time points and subjected to sodium bisulfite conversion, followed byMassARRAY to detect DNA methylation frequencies. The amplicon I is located −1069 to − 623 bps upstream of the translation start site. Eachcircle represents a CpG dinucleotide. Color code: yellow=unmethylated CpG, blue= 100% methylated CpG. Gray circles indicate not analyzedmethylation values due to CpGs with high or low mass Dalton peaks falling outside the conservative window of reliable detection for theEpiTYPER software.


5432


demonstrated that both proteins catalyzed the incorporation ofDNAme and induced potent cell growth arrest. This effect wasdependent on both a functional DBD and the catalytic function ofDNMT3A, as an untargeted DNMT3A construct or the expressionof a catalytic mutant resulted in an unspecific backgroundincorporation of DNAme, which was not capable of silencingSOX2 or inhibit cell growth. Furthermore, de novo DNAme wasstably maintained in vivo even 53 days after removal of the ZF598-DNMT3A and was accompanied with a sustained suppression ofSOX2 expression and tumor growth inhibition. It is important tonote that the absence of expression of the methyltransferase

upon long-term removal of Dox was carefully validated usingimmunoblotting and immunofluorescence.In this study, we utilized modular ZF proteins engineered to

bind an 18-bp sequence in the core promoter of SOX2. ZFs arewell-characterized DBDs and have been used for almost twodecades for DNA targeting with customizable sequenceselectivity.27 Furthermore, recent novel approaches to targetendogenous gene expression also hold great promise such astranscription activator-like effectors28,29 and the RNA-guidedclustered regularly interspersed short palindromic repeats (Cas)system.30,31 Both transcription activator-like effectors and Cas9

empty vector+Dox

ZF598-DNMT3Ano Dox

ZF598-DNMT3A+Dox

empty vectorno Dox

Ki67

SOX2

nuclear

HA-tag

empty vector+Dox

ZF598-DNMT3A +Dox

ZF598-DNMT3ADox-removal

20x 20x 20x

ZF598-DNMT3ADox-removal

Figure 5. Histological and immunofluorescence analyses of tumor sections reveals phenotypic memory. (a) Hematoxylin and Eosin stains ofrepresentative ZF598-DNMT3A −Dox, +Dox and Dox-removal tumor sections. Sections of empty vector tumors were extracted 29 days postinduction, sections of ZF598-DNMT3A +Dox were sampled after 54 days of Dox induction, and those of Dox-removal samples were harvested45 days after Dox removal. Pictures were taken at × 20 and a detail of the image is shown. (b) Immunofluorescence on sections of emptyvector and ZF598-DNMT3A tumors collected 29 days post induction. The expression of ZF598-DNMT3A (HA-tag, green), SOX2 (red) and theproliferation marker Ki-67 (green, bottom) in +Dox, −Dox and Dox-removal conditions are indicated. Images are taken at × 40 magnification.


5433


could be similarly used as alternatives to ZFs to target the catalyticactive domain of DNMT3A to specific chromosomal sites.Pioneering work has been recently published by Konermann

et al.32 who fused 32 different histone effector domains to atranscription activator-like effector DBD targeting the Neurog2locus and demonstrated transcriptional repression. However,the spatio-temporal dynamics associated with de novo DNAmeand histone post-transcriptional modifications still remainelusive. Here we demonstrate that de novo DNAme is faithfullytransmitted after clearance of expression of the 6ZF-DNMT3Afusions. The artificial incorporation of DNAme provides aninitial platform, which is read, written and propagated byendogenous methylation machinery during DNA replication. Thecomplex between DNMT1, UHRF1 (ubiquitin-like PHD (containingplant homeodomain) and RING (really interesting new gene)finger domains 1) and PCNA (proliferating cell nuclear antigen) atthe replication fork could mediate the cross-talk betweenmethylated DNA and repressive histone modifications.33,34 In ourexperiments, it is conceivable that such mechanisms could beresponsible for the maintenance of DNAme and the down-regulation of SOX2.35

In addition to hereditable transmission during mitosis, weobserved that the de novo methylation in the SOX2 locus wasexpanded at least 1 Kbp away from the targeted site. Potentialmechanisms could involve the endogenous spreading of DNAmefor example by chromatin looping or iterative reading and writingof DNA and histone post-transcriptional modifications,36 or bylow-frequency occupancy of these sites by the 6ZF domains. Inthis regard, it will be interesting to investigate whether this effectis also observed with more recently developed engineeringapproaches such as transcription activator-like effectors orclustered regularly interspersed short palindromic repeats/dCas9.In contrast with DNMT3A, cells expressing the ZF-SKD fusions

did not sustain a stable downregulation of SOX2, indicatingmechanistically distinct epigenetic processes initiated by theZF-DNMT3A and ZF-SKD constructs. SKD mediates silencing oftarget genes through recruitment of KAP1 (Krueppel-associatedprotein 1), which acts as a scaffold for heterochromatin-inducingmodifiers such as HP1 and methyltransferase SETDB1.36 However,unlike DNMT3A, SKD has no intrinsic enzymatic activity, andmainly carries a recruiting activity. In our study, the suppression ofSKD expression did not result in epigenetic memory.To date, four epigenetic drugs have been approved by the US

Food and Drug Administration, including two DNMT and twoHDAC inhibitors37 to reactivate aberrantly silenced tumorsuppressor genes. Here we report a novel strategy that willenable the silencing of aberrantly expressed oncogenes. We havepreviously demonstrated systemic delivery of lipid-protamine-RNAnanoparticles encapsulating mRNA encoding a ZF proteinupregulating the MASPIN promoter for the treatment of serousovarian cancer.38 Such lipid-protamine-RNA technology couldbe similarly adapted to deliver the ZF-DNMT3A constructs forstable heterochromatization of SOX2. The small, compactmolecular architecture of ZF domains and their lack of immuno-genicity make them very suitable molecular scaffolds for this typeof delivery.In summary, we demonstrate the applicability of DBD-DNMT3A

fusions to induce targeted DNAme to stably repress oncogenicexpression in a long-term xenograft mouse model of breastcancer. This approach could be extended to many otheroncogenic drivers for which no drug is currently available topromote potent and durable cancer cell growth inhibition. Inaddition to its important implications in cancer therapeutics, ourapproach provides a platform to induce targeted DNAme andinvestigate the temporal and spatial propagation of this epige-netic state and its effects on gene expression on a genomic levelin vivo.

MATERIALS AND METHODSGeneration of stable cell linesThe construction of the SKD, DNMT3A, DNMT3A-E74A and the 6ZFdomains has been described elsewhere.17,19 ZF598-DNMT3A and ZF598-DNMT3A-E74A were cloned into pRetroX-Tight-Pur (CloneTech, MountainView, CA, USA). Generation of MCF7 cells stably expressing ZF598-DNMT3Aand ZF598-DNMT3A-E74A was performed as described.17 Cells wereinduced with Dox every 48 h and either harvested at 72 h after the firstinduction (+Dox) or removed from Dox and subcultured for 8 days andprocessed for western blot.

Cell proliferation assaysEighteen replicates of MCF7 cells stably expressing empty vector, ZF598-SKD and ZF598-DNMT3A were plated in 96-wells plates (1000 cells/well).Twelve replicates were induced with Dox at time point 0 and after 48 h.After 72 h, six replicates were removed from Dox, whereas six replicateswere continuously induced. Cell proliferation was assessed by CellTiterGloassay (Promega, Madison, WI, USA) every 24 h for a total period of 144 h.39

Results were normalized to readings at day 0.

Bisulfite conversion and MassARRAY methylation analysisAfter genomic DNA extraction, 2 μg of sample DNA (derived from eithercell line or tumor) was treated with sodium bisulfite using the EZ DNAMethylation-Direct Kit (Zymo Research, Irvine, CA, USA). We customdesigned primers for three amplicons spanning the core SOX2 promoter,one specifically including the 6ZF-binding site 5′-gGCCCCCTCCTCCCCCGGC-3′ and two amplicons up- and downstream of the 6ZF-bindingsite. Polymerase chain reaction was then carried out on 5–10 ng of sodiumbisulfite converted sample DNA using conversion-specific primers. Foramplicon I: forward primer 5′-aggaagagagGGATAGAGGTTTGGGTTTTTTAATTT-3′ and reverse primer 5′-cagtaatacgactcactatagggagaaggctAAACCAACCTACCAACCACTAAAA-3′. For amplicon II: forward primer 5′-aggaagagagAAAGGTTTTTTAGTGGTTGGTAGGT-3′ and 5′-agtaatacgactcactatagggagaaggctAAAACTCAAACTTCTCTCCCTTTCT-3′ reverse primer. For ampliconIII: 5′-aggaagagagTTTTGGTATGGTTTTTGGTTTTATG-3′ forward primerand 5′-cagtaatacgactcactatagggagaaggctAATTTTCTCCATACTATTTCTTACTCTCC-3′ reverse primer with lower case letters representing the T7 tagsequences. Percent SOX2 DNAme was quantified using mass spectrometrywith the SEQUENOM EpiTYPER T complete reagent kit (San Diego, CA,USA). Conventional sodium bisulfite-sequencing analysis was carried out asdescribed before using 5′-AAAGGTTTTTTAGTGGTTGGTAGGT-3′ forwardprimer and 5′-AAAACTCAAACTTCTCTCCCTTTCT-3′ reverse primer forpolymerase chain reaction amplification of bisulfite converted DNA.19

Primers to detect the % methylation for the MASPIN amplicon using theEpiTYPER platform: forward primer 5′-aggaagagagGAGGTTTTTTGGAAGTTGTGTAGAT-3′ and reverse primer 5′-cagtaatacgactcactatagggagaaggctCCCCACCTTACTTACCTAAAATCAC-3′

Mouse experimentsFemale NUDE mice (age 4 weeks) were purchased from Taconic Farms(Hudson, NY, USA) and housed under pathogen-free conditions. TheInstitutional Animal Care and Use Committee at the University of NorthCarolina at Chapel Hill approved all experiments described herein.Estrogen pellets containing 2 mg 17β-Estradiol (Sigma-Aldrich Corp.,St Louis, MO, USA) and 8mg Cellulose (Sigma-Aldrich Corp.) weresubcutaneously implanted in the animals 7 days prior of the injection ofthe cells. MCF7 cells (2 × 106) were collected and resuspended withmatrigel (BD Bioscience, San Diego, CA, USA) 1:1 volume ratio in a totalvolume of 100ml. The cell–matrigel mixture was injected into the mouseflank of N=11 mice for empty vector and N=22 for ZF598-DNMT3A.Tumor growth was monitored by caliper twice a week. When the tumorreached a size of ~ 25–30mm3, Dox was administered to the mice in theform of green food pellets (200mg/kg of mice chow) for a period of19 days. At day 19 post induction half of the animals of each group wereremoved from Dox, whereas the other half was maintained under Dox diet.During the entire experiment, the mice weight was monitored to ensureabsence of toxicity. Animals were killed when tumor reached 100mm3

(empty vector 29 days post induction, ZF598-DNMT3A upon Dox removal(‘No Dox’) for 43 days post induction and ZF598-DNMT3A Dox removal at72 days post induction). Statistical differences between control andZF-DNMT3A animals were assessed by Wilcoxon Ranks Sum Test analysis.


5434


CONFLICT OF INTERESTThe authors declare no conflicts of interest.

ACKNOWLEDGEMENTSThis research was supported by an Australian Research Council (ARC) FutureFellowship FT130101767, the Cancer Council of Western Australia Research Fellow-ship, National Health and Medical Research Council grant APP1069308, NationalInstitutes of Health grants R01CA170370, R01DA036906, the National Breast CancerFoundation NC-14-024 (PB) and the Barbara Senich Genomics Innovation Award, theBarbara Senich Genomics Innovation Endowment Fund and the Susan G KomenFoundation KG090180 (TSS).

REFERENCES1 Kouzarides T. Chromatin modifications and their function. Cell 2007; 128:

693–705.2 Beisel C, Paro R. Silencing chromatin: comparing modes and mechanisms. Nat Rev

Genet 2011; 12: 123–135.3 Lange UC, Schneider R. What an epigenome remembers. Bioessays 2010; 32:

659–668.4 Sandoval J, Esteller M. Cancer epigenomics: beyond genomics. Curr Opin Genet

Dev 2012; 22: 50–55.5 The Cancer Genome Atlas Network. Comprehensive molecular portraits of human

breast tumours. Nature 2012; 490: 61–70.6 Stephens PJ, Tarpey PS, Davies H, Van Loo P, Greenman C, Wedge DC et al. The

landscape of cancer genes and mutational processes in breast cancer. Nature2012; 486: 400–404.

7 Curtis C, Shah SP, Chin SF, Turashvili G, Rueda OM, Dunning MJ et al. The genomicand transcriptomic architecture of 2,000 breast tumours reveals novel subgroups.Nature 2012; 486: 346–352.

8 Boyer LA, Lee TI, Cole MF, Johnstone SE, Levine SS, Zucker JP et al. Coretranscriptional regulatory circuitry in human embryonic stem cells. Cell 2005; 122:947–956.

9 Brazel CY, Limke TL, Osborne JK, Miura T, Cai J, Pevny L et al. Sox2 expressiondefines a heterogeneous population of neurosphere-forming cells in the adultmurine brain. Aging Cell 2005; 4: 197–207.

10 Sikorska M, Sandhu JK, Deb-Rinker P, Jezierski A, Leblanc J, Charlebois C et al.Epigenetic modifications of SOX2 enhancers, SRR1 and SRR2, correlate within vitro neural differentiation. J Neurosci Res 2008; 86: 1680–1693.

11 Sussman RT, Stanek TJ, Esteso P, Gearhart JD, Knudsen KE, McMahon SB. Theepigenetic modifier ubiquitin-specific protease 22 (USP22) regulates embryonicstem cell differentiation via transcriptional repression of sex-determining regionY-box 2 (SOX2). J Biol Chem 2013; 288: 24234–24246.

12 Schmitz M, Temme A, Senner V, Ebner R, Schwind S, Stevanovic S et al.Identification of SOX2 as a novel glioma-associated antigen and potential targetfor T cell-based immunotherapy. Br J Cancer 2007; 96: 1293–1301.

13 Bareiss PM, Paczulla A, Wang H, Schairer R, Wiehr S, Kohlhofer U et al. SOX2expression associates with stem cell state in human ovarian carcinoma. Cancer Res2013; 73: 5544–5555.

14 Bass AJ, Watanabe H, Mermel CH, Yu S, Perner S, Verhaak RG et al. SOX2 is anamplified lineage-survival oncogene in lung and esophageal squamous cellcarcinomas. Nat Genet 2009; 41: 1238–1242.

15 Alonso MM, Diez-Valle R, Manterola L, Rubio A, Liu D, Cortes-Santiago N et al.Genetic and epigenetic modifications of Sox2 contribute to the invasive pheno-type of malignant gliomas. PLoS ONE 2011; 6: e26740.

16 Chen Y, Shi L, Zhang L, Li R, Liang J, Yu W et al. The molecular mechanismgoverning the oncogenic potential of SOX2 in breast cancer. J Biol Chem 2008;283: 17969–17978.

17 Stolzenburg S, Rots MG, Beltran AS, Rivenbark AG, Yuan X, Qian H et al. Targetedsilencing of the oncogenic transcription factor SOX2 in breast cancer. NucleicAcids Res 2012; 40: 6725–6740.

18 Xiang R, Liao D, Cheng T, Zhou H, Shi Q, Chuang TS et al. Downregulation oftranscription factor SOX2 in cancer stem cells suppresses growth and metastasisof lung cancer. Br J Cancer 2011; 104: 1410–1417.

19 Rivenbark AG, Stolzenburg S, Beltran AS, Yuan X, Rots MG, Strahl BD et al.Epigenetic reprogramming of cancer cells via targeted DNA methylation.Epigenetics 2012; 7: 350–360.

20 Siddique AN, Nunna S, Rajavelu A, Zhang Y, Jurkowska RZ, Reinhardt R et al.Targeted methylation and gene silencing of VEGF-A in human cells by using adesigned Dnmt3a-Dnmt3L single-chain fusion protein with increased DNAmethylation activity. J Mol Biol 2013; 425: 479–491.

21 Li F, Papworth M, Minczuk M, Rohde C, Zhang Y, Ragozin S et al. Chimeric DNAmethyltransferases target DNA methylation to specific DNA sequences andrepress expression of target genes. Nucleic Acids Res 2007; 35: 100–112.

22 de Groote ML, Verschure PJ, Rots MG. Epigenetic Editing: targeted rewriting ofepigenetic marks to modulate expression of selected target genes. Nucleic AcidsRes 2012; 40: 10596–10613.

23 Nunna S, Reinhardt R, Ragozin S, Jeltsch A. Targeted methylation of the epithelialcell adhesion molecule (EpCAM) promoter to silence its expression in ovariancancer cells. PLoS ONE 2014; 9: e87703.

24 Chaikind B, Ostermeier M. Directed evolution of improved zinc finger methyl-transferases. PLoS ONE 2014; 9: e96931.

25 Blancafort P, Jin J, Frye S. Writing and rewriting the epigenetic code of cancercells: from engineered proteins to small molecules. Mol Pharmacol 2013; 83:563–576.

26 Stolzenburg S, Bilsland A, Keith WN, Rots MG. Modulation of gene expressionusing zinc finger-based artificial transcription factors. Methods Mol Biol 2010; 649:117–132.

27 Choo Y, Sanchez-Garcia I, Klug A. In vivo repression by a site-specific DNA-bindingprotein designed against an oncogenic sequence. Nature 1994; 372: 642–645.

28 Boch J, Scholze H, Schornack S, Landgraf A, Hahn S, Kay S et al. Breaking the codeof DNA binding specificity of TAL-type III effectors. Science 2009; 326: 1509–1512.

29 Bogdanove AJ, Voytas DF. TAL effectors: customizable proteins for DNA targeting.Science 2011; 333: 1843–1846.

30 Jinek M, Chylinski K, Fonfara I, Hauer M, Doudna JA, Charpentier E. A program-mable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity.Science 2012; 337: 816–821.

31 Wiedenheft B, Sternberg SH, Doudna JA. RNA-guided genetic silencing systems inbacteria and archaea. Nature 2012; 482: 331–338.

32 Konermann S, Brigham MD, Trevino A, Hsu PD, Heidenreich M, Le C et al. Opticalcontrol of mammalian endogenous transcription and epigenetic states. Nature2013.

33 Fuks F, Burgers WA, Brehm A, Hughes-Davies L, Kouzarides T. DNA methyl-transferase Dnmt1 associates with histone deacetylase activity. Nat Gen 2000; 24:88–91.

34 Esteve PO, Chin HG, Smallwood A, Feehery GR, Gangisetty O, Karpf AR et al. Directinteraction between DNMT1 and G9a coordinates DNA and histone methylationduring replication. Genes Dev 2006; 20: 3089–3103.

35 Rodriguez J, Munoz M, Vives L, Frangou CG, Groudine M, Peinado MA. Bivalentdomains enforce transcriptional memory of DNA methylated genes incancer cells. Proc Natl Acad Sci USA 2008; 105: 19809–19814.

36 Hathaway NA, Bell O, Hodges C, Miller EL, Neel DS, Crabtree GR. Dynamics andmemory of heterochromatin in living cells. Cell 2012; 149: 1447–1460.

37 Heyn H, Esteller M. DNA methylation profiling in the clinic: applications andchallenges. Nat Rev Genet 2012; 13: 679–692.

38 Lara H, Wang Y, Beltran AS, Juarez-Moreno K, Yuan X, Kato S et al. Targetingserous epithelial ovarian cancer with designer zinc finger transcription factors.J Biol Chem 2012; 287: 29873–29886.

39 Beltran AS, Rivenbark AG, Richardson BT, Yuan X, Quian H, Hunt JP et al.Generation of tumor initiating cells by exogenous delivery of OCT4Transcription Factor. Breast Cancer Res 2011; 13: R94.

This work is licensed under a Creative Commons Attribution 4.0International License. The images or other third party material

in this article are included in the article’s Creative Commons license, unlessindicated otherwise in the credit line; if the material is not included under the CreativeCommons license, users will need to obtain permission from the license holder toreproduce thematerial. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)


5435


http://creativecommons.org/licenses/by/�4.0/

http://creativecommons.org/licenses/by/�4.0/

stable oncogenic silencing in vivo by programmable and ......stable oncogenic silencing in vivo by...

Documents